>It's on the luddites to demonstrate evidence of harm, if they want some use of technology banned.
Here's one study from last year (1000 students) [1]. They showed that improved outcomes with AI are short term, and likely a result of AI doing the work for the student. Their methodology involved a test without the use of AI at a later date, and one of the groups who used AI tested much worse than the control group who did not use AI during learning.
Moreover, a tertiary group that used a specially trained AI for tutoring showed no improvement in testing outcomes over the control group.
[1] Bastani, H., Bastani, O., Sungu, A., Ge, H., Kabakçı, Ö., & Mariman, R. (2025). "Generative AI Without Guardrails Can Harm Learning: Evidence from High School Mathematics." PNAS, 122. PNAS
How do you ensure the kid is learning the right material? You want them to stay on topic, but a Socratic Method that starts off on one topic can easily branch into many other topics, and then suddenly your student has spent the whole class learning about something other than what they were supposed to.
All learning is good, yes. But a baseline curriculum is necessary to create a ground floor of knowledge for society.
>I can assure you that no matter what concerns you may have about hallucinations in LLMs, I can bet everything I have that a reasonably modern model (and I'm thinking in the range of gemini-flash, not Fable) in a well designed harness geared towards tutoring would handily and repeatedly outperform every single teacher I had all throughout my schooling.
What are the *social* ramifications of an education system where children learn increasingly from AI tutors, while staring at a screen in a classroom for multiple hours a day?
Children need to spend time interacting with other children, as well as with adults. We have a mountain worth of data that children have become addicted to their devices, why are we doubling down on this by including MORE screen time at such an early age?
>If implemented right
This same argument could be applied to any technology, or policy that has ever been created. But we don't live in a world where people can do things the right way because of any number of externalities that muddies the actual implementation.
>Another angle to explore is how much of the process of software development that we do manually can we automate, particularly the parts that still require human input (like code reviews). That may also help with reducing the cognitive load.
Get your vibe coded software with zero human reviewers away from the public infrastructure.
Nothing more than a bunch of people who haven't actually tried writing, and therefore aren't aware of what good writing actually looks like.
This is the problem with LLMs. It allows neophytes to trick themselves into believing that they're now a writer/programmer/artist for prompting a model, and because they don't know what they don't know about writing/programming/art, they think it's good when it's actually slop.
>AI slop is unsatisfying because there is no there there. It is intellectual junk food that mimics nutrition but delivers only empty calories. Satisfying AI outputs must embed dense information and compute to actually reward a reader's attention. You inject this value through brute-force search, non-trivial prompting, and rigorous curation, ensuring the final result reflects genuine algorithmic effort rather than the zero-shot 'WYSIWYG' default.
I don't agree with this at all. You can prompt all you want, but if you don't have the actual skills to make good art, you won't really know if what you're generating is good or not. This is to say nothing of the "hollow-ness" of outsourcing your voice to generative algorithms.
What is "high literature"? Have you actually read any of the greats? I have, and while I'm not a fan of everything I've read, I never felt inundated with constant metaphors and overly eloquent prose.